I'll keep this technical and skip the feelings part.
BCG's engineering pipeline includes a HackerRank OA and then two live coding rounds in the onsite. Here's what I observed:
Online Assessment (HackerRank) 90 minutes 3 problems in my case: one easy warmup (string manipulation), one medium (tree traversal), one that was technically medium but had an edge case that would catch you if you just pattern-matched to the obvious solution No webcam, no proctor for the one I did Standard HackerRank judge, test cases visible on pass, hidden on fail
The third problem is where they're probably differentiating. I finished with about 10 minutes to spare but nearly got the third one wrong because I didn't read the constraints carefully. The input size meant a naive O(n^2) would time out on large cases but pass the visible tests. Classic.
Live coding rounds (2 rounds, 45 min each) Done in a shared editor, I used Python First round: graph problem, BFS/DFS variant, interviewer gave hints when I got stuck but tracked how fast I needed them Second round: more of a design-and-code hybrid, less algorithms, more about structuring a small feature cleanly
I'd call both rounds "LeetCode medium" in difficulty. Nothing I'd classify as hard. But they were watching how I worked, not just whether I got the right answer. I talked through my approach before coding anything, which the interviewer explicitly said was helpful.
Couple of notes: They do not require you to memorize sorting algorithms from scratch. I was not asked to implement quicksort. What they cared about was: do you know when to use a hashmap vs. a heap, and can you reason about complexity.
If you're coming from a pure consulting background and trying to get into BCG's tech org, the OA is the real filter. Brush up on trees, graphs, and hashing. Not DP-heavy in my experience.